Access and Utilization of New Antidepressant and Antipsychotic Medications

As health care in the United States has become increasingly cost-conscious, providers are increasingly encouraged not only to consider the outcomes of treatment alternatives, but also to consider and justify costs. Similarly, payers have begun to examine the interaction of cost and effectiveness more carefully. The cost-effectiveness of newer antidepressants^44-49 and antipsychotics^50-52 has been the subject of several recent reviews. In addition, numerous reviews of individual agents exist.^53-58 The following review of the cost-effectiveness literature is designed to provide a high-level overview of the subject and the principal conclusions. The reader is referred to the reviews cited above for more detailed information.

In brief, the following limited review of the literature found:

Antidepressants

The literature provides no consistent differences in total treatment costs associated with the use of different antidepressant agents, although individual studies have claimed that one particular agent is superior to another.

• The vast majority of studies are sponsored by pharmaceutical manufacturers.
• Most of these studies have found that treatment with the pharmaceutical manufactured by the sponsor is no more costly than treatment with comparative agents.
• It is unlikely that manufacturers would allow publication of a study that found a negative result for their agent.
• However, results from studies sponsored by manufacturers do not differ substantially from those obtained from studies sponsored by independent or government sources.

• Antidepressant treatment with fluoxetine may yield lower costs than treatment with TCAs -- particularly for post-initiation therapy. Stronger evidence exists, however, for greater compliance with fluoxetine than with TCAs.
• Studies on the costs of paroxetine relative to imipramine are inconclusive; however the evidence is strong for greater compliance or treatment continuation for paroxetine.
• Treatment with sertraline may result in lower total costs than treatment with TCAs. One study also asserts that sertraline is more effective than TCAs. Taken together, it is reasonable to conclude that treatment with sertraline is more cost-effective than treatment with TCAs.
• The evidence on venlafaxine tends to suggest that treatment with venlafaxine is more cost-effective than treatment with either TCAs or trazadone.
• Although two studies differ on the relative costs of treatment with nefazodone, both studies do suggest that treatment with nefazodone is more cost-effective than treatment with imipramine.

Evidence from the pharmacoeconomic literature shows that treatment using the new antipsychotics exhibits economic advantages over treatment using the older antipsychotics. This evidence for olanzapine and clozapine is superior to the corresponding evidence for risperidone.

• The limited models comparing risperidone and clozapine to the older antipsychotics suggest that, despite their higher medication costs, risperidone and clozapine may be more cost-effective than older antipsychotics (e.g., haloperidol) in the treatment of schizophrenia.
• The pharmacoeconomic evidence on risperidone versus typical antipsychotics presents no clear advantage for risperidone in terms of efficacy; although on balance, risperidone may have a slight edge in cost-effectiveness.
• Olanzapine appears to be more cost-effective than haloperidol in the treatment of schizophrenia.
• The manufacturers of quetiapine have carefully designed a naturalistic comparative pharmacoeconomic trial in order to compare quetiapine to usual care. No results are yet available from this trial.
• The results of numerous studies together provide strong evidence of clozapine's superior cost-effectiveness relative to older, typical antipsychotics in treatment-resistant schizophrenia.

• Current inconsistencies regarding the dosages in the 2^nd generation atypical antipsychotics head-to-head (risperidone vs. olanzapine) studies render the evidence inconclusive as to which agent is more cost-effective.

A. Overview of Pharmacoeconomics

The term "cost-effectiveness" has been broadly applied to several methods of pharmacoeconomic analysis which compare the clinical and/or quality of life outcomes and costs associated with competing pharmacological agents. Measurements used in pharmacoeconomic analyses of the newer psychotropic therapies include cost-of-illness, cost-minimization, cost-effectiveness, cost-benefit, and cost-utility. The principal difference among these types of analyses are outlined in Exhibit VII-1 below.

In addition to differences in analysis strategy, pharmacoeconomic studies vary by method of data collection. Although stand-alone medical outcomes or economic measurements collected during randomized clinical trials (i.e., "piggyback" trials) are the "gold-standard" for pharmacoeconomic analyses, these are expensive to conduct and have not been widely implemented in the study of mental health. Retrospective or prospective medical claims analysis and decision-analytic economic models are far more common.⁵⁹

To date, few well-designed pharmacoeconomic studies have been conducted to evaluate the cost-effectiveness of the newer antipsychotic and antidepressant medications compared to conventional therapies. Although experts emphasize the superiority of large-scale, prospective, comparative studies in pharmacoeconomic evaluation,⁵⁹ the majority of studies use less reliable, and perhaps more convenient methods of economic analysis.

Analysis of claims from a health care payer can take one of several approaches. If the comparison groups are used concurrently in clinical practice, it may be possible to identify patients who have taken either therapy and analyze average costs between treatment groups. The majority of economic studies to date in the literature are uncontrolled retrospective historical analyses which estimate cost-savings for competing treatments. In these studies the researchers examine the costs of treatment (and other outcomes) one year prior to and one to two years following treatment with the study medication. In the pre-study phase, patients may have been treated with one of several therapeutic strategies. For this reason, such studies lack a true control group, making it difficult to attribute cost and clinical outcomes to a specific course of therapy.

Decision-analytic models are a convenient tool to estimate the true cost-effectiveness of treatment over a lifetime. However, these models rely heavily on clinical and economic assumptions because good evidence on lifetime costs and utilities are not currently available.⁵³ Sensitivity analyses that adjust the clinical and economic outcomes probabilities are essential to demonstrating the validity of these models. This is especially true when one outcome parameter may be ill-defined or where these parameters may be obtained from studies that are not strictly applicable to a real-world setting.

Until the mid to late 1990's, retrospective, often uncontrolled, historical analyses were most common in evaluating cost-effectiveness of newer antipsychotics, whereas retrospective reviews of health care payer claims were most common for evaluating cost-effectiveness of newer antidepressants. A small number of large-scale real-world effectiveness trials have recently been conducted or are currently in progress for both antidepressants and antipsychotics.

In the 13 years since the initial launch of bupropion in the United States, physicians and patients have rapidly adopted new-generation antidepressant and antipsychotic medications as first-line treatments for depressive and psychotic illnesses. These new agents have provided treatment alternatives to established therapies such as the tricyclic antidepressants (TCAs) and the typical antipsychotics that had been in use for more than two decades. These new agents have also brought with them price tags that are substantially larger than those associated with the agents they hoped to replace.

Although it has been broadly assumed in the popular understanding that new generation antidepressant and antipsychotic medications are more effective than the older generations, the argument for their use, based on efficacy alone, is far from clear. A recent meta-analysis of antidepressant agents published by the Agency for Health Care Policy and Research concluded that SSRIs and other new-generation antidepressants are no more effective than TCAs.⁶⁰ As a result, the rationale for the use of these newer (and more expensive) agents has rested on arguments surrounding ease of use, tolerability, general safety, safety in overdose (i.e., easy of suicide attempts), and (perhaps consequently) net cost-savings to the total health care system. Indeed, Hylan and coworkers warn that compliance with therapy by a patient depends on the particular pharmaceutical used itself, and therefore assuming equal efficacy as a result of equal use is not tenable.⁴⁸

B. The Cost-Effectiveness of Newer Antidepressant Medications

1. The Cost of Depressive Illness in the United States

   Any understanding of the cost-effectiveness of a medical therapy must be based on both the cost of the illness being treated as well as the effectiveness of the therapy. Several cost of illness studies on depressive or affective disorders have been conducted in the US. One estimate placed the cost of affective disorders in 1990 at $30.4 billion, an increase of 46% over 1985.⁶¹ Medications accounted for approximately 2% of the $406 million direct medical costs in this study. These researchers also estimated that in 1985, the U.S. lost $6 billion in indirect costs, largely due to deaths related to these disorders.

   Another study estimated that the total cost of depressive disorders in the U.S. was $43.7 billion. ⁶² Of this amount, direct medical costs accounted for more than $12 billion, with $7.5 billion lost to indirect costs. The cost of medications was estimated at $1.17 billion, or approximately 10% of direct medical costs. Note, however, that both of these studies were conducted before the introduction of most of the antidepressant agents of interest in this report.

2. Pharmacoeconomic Analyses of Specific Pharmaceutical Agents

   The Lewin Group identified 40 studies that evaluate the pharmacoeconomics of newer antidepressant medications to a reliable level of rigor. Twenty-nine studies compare newer antidepressants to older antidepressants; 11 studies compare newer antidepressants to other newer antidepressants. The majority of these studies are retrospective analyses of claims data from health care payers, predominantly managed care plans. Several other studies are decision-analytic models that incorporate both claims data, the results of randomized clinical efficacy trials, and the results of expert panel interviews. Finally, a small number of studies are prospective, randomized economic clinical trials conducted either in a naturalistic health care setting or as a "piggy back" to a clinical trial.

   On balance, the literature shows that there is no overall difference in total treatment costs associated with the use of different pharmaceutical agents, although individual studies have claimed that one particular agent is superior to another. In some cases, studies have found that treatment with newer agents is cost-saving compared to treatment with older agents, whereas in others the reverse has been shown to be true. In numerous cases, researchers have failed to find a difference between treatment groups.

   The majority of these studies have been sponsored by pharmaceutical manufacturers. As a result, one might expect publication bias to color the results. In fact, in all the studies reviewed below, treatment with the agent manufactured by the sponsor appears to be at least no more costly in terms of total direct medical costs than treatment with the comparative agents. The claim that one newer agent is economically superior over another may appear inherently specious if the study is sponsored by the manufacturer claiming superiority. However, the result that treatment with new-generation antidepressants is at worst no more costly in terms total direct medical costs than treatment with older agents is confirmed by studies sponsored by government agencies or other third parties.

   The following discussion is limited to studies conducted from the perspective of the US healthcare system. Reference is made to studies conducted in other countries when they are of very high quality or the results compare and contrast to those for the US.

   1. Newer Antidepressants Compared to Older Antidepressants

      1. SSRIs as a Class

         Smith and Sherrill found that cost of treatment of depressed patients in TennCare with SSRIs or with TCAs did not differ substantially.⁶³ This retrospective 12-month analysis of claims from a U.S. Medicaid agency (TennCare) compared total costs incurred 76 patients treated with an SSRI (fluoxetine, paroxetine, or sertraline) and 76 patients treated with a TCA. All patients had a confirmed diagnosis of major depressive disorder (ICD-9 code 296.2, 296.3) or depression not elsewhere classified (ICD-9, 311). The total cost for depression-related treatment for a patient treated with an SSRI was $54,164, whereas for those treated with a TCA this cost was $55,838. Because this difference did not achieve statistical significance, the authors concluded that the cost of treating depression was identical between the two treatment groups. It is worth noting that patients covered by the Medicaid agency whose claims were used had been required to experience two treatment failures prior to initiating SSRI therapy. It is therefore possible that the SSRI patients in this study were inherently more difficult to treat than were the TCA patients.

         Boyer and Feigner found that treatment of depressed patients with an SSRI led to substantially lower costs than treatment with a TCA when both short-term and maintenance costs were considered.⁶⁴ Total costs were lower even though total pharmaceutical costs were much higher for the SSRI group. Their model compared direct costs over a six-month period of initial treatment with an SSRI (fluoxetine, paroxetine, or fluvoxamine) versus initial treatment with a TCA. The model assumed that 10% of patients with acute depression would require maintenance treatment, and that 10% would withdraw. The model also assumed that 65% of those who withdrew and 25% of those who were compliant would experience relapse. Based upon these assumptions, the investigators concluded that initial treatment with TCAs would be $1,374 less than with SSRIs. However, when both short-term and maintenance therapy costs were considered, SSRIs were less expensive than TCAs. Per 1000 patients, direct costs of starting treatment with an SSRI, costs of hospitalization, and costs of outpatient visits were estimated to be $47,257, $67,188, and $11,546 less, respectively, than with TCAs. Drug costs were estimated to be $31,476 per 1000 patients more for SSRIs than for TCAs.

      2. Fluoxetine

         In pharmacoeconomic studies comparing an SSRI to TCAs, fluoxetine has been the most frequently studied antidepressant. Some of these studies also have arms comparing fluoxetine to other SSRIs. Only one of these studies is a randomized prospective trial conducted in a naturalistic setting.⁴⁶

         Taken together, this set of studies suggests that treatment with fluoxetine may yield lower costs than treatment with TCAs -- particularly for post-initiation therapy. Most studies conclude there is no difference between the total costs for treating a patient with fluoxetine and the total costs for treating a patient with a TCA, although a few claim economic superiority for fluoxetine over TCAs. Stronger evidence exists, however, for greater compliance with fluoxetine than with TCAs.

         Simon and Fishman⁴⁶ conducted the only prospective, randomized economic clinical trial comparing the cost of treating depressed patients with fluoxetine first-line versus the cost of treating with desipramine or imipramine in a primary care HMO setting. In an intent-to-treat analysis of the six-month trial, the investigators observed no significant difference among drug groups for symptom reduction, Quality of Life, or total healthcare cost endpoints. Patients initiating on fluoxetine were more likely to continue the originally prescribed antidepressant and were also more likely to refill prescriptions than were patients initiating on a TCA. Furthermore, the dropout rate during the first month of treatment was higher in the TCA groups than in the fluoxetine groups. Dropouts were largely driven by unacceptable adverse effects.

         Croghan et al. conducted a retrospective database study using 2-stage econometric modeling to estimate healthcare costs of primary care patients receiving one year of initial therapy with fluoxetine (n=799), trazadone (n=89), or one of two classes of TCAs (amitriptyline/imipramine (n=104) or desipramine/nortriptyline (n=250)). Overall, the results suggest that fluoxetine is at least as cost-effective as trazadone or the TCAs.⁶⁵ Total direct healthcare costs per patient (including hospitalizations, labs, radiology, physician visits, other outpatient services and drugs costs) were no different between the four treatment groups, although total mental healthcare costs were less for the trazadone cohort than for fluoxetine or the TCAs. However, the percentage of patients continuing medication continuously for at least six months was substantially greater with fluoxetine than with trazadone and the TCAs. Likewise, the percentage of patients switching from initial therapy was lowest with fluoxetine. Although the study is limited by its homogenous population, it did control for observed and unobserved characteristics in the econometric modeling, adding credence to the results.

         Sclar et al. conducted a retrospective database study coupled with a multivariate regression analysis to compare the total one-year direct medical costs accumulated by patients prescribed either fluoxetine or one of three TCAs (amitriptyline, nortriptyline, desipramine).⁶⁶ The researchers concluded that use of a TCA was associated with greater per patient post-initiation total healthcare costs ($366) compared to treatment with fluoxetine. The drivers of this increased cost included increases in physician visits, hospital visits, and laboratory costs. Although this study included no direct measure of treatment outcomes, it did show that patients were more likely to continue treatment on fluoxetine than they were on a TCA. The 701 patients in the study were representative of private sector depressed patients. The average age was over 40 and approximately 75% were women. This study, however, did not control for unobserved differences across patients.

         In a later study, Sclar, et al. compared claims of patients treated with fluoxetine (n=180) to those treated with amitriptyline (n=211) and nortriptyline (n=159).⁶⁷ This study tracked costs over a 3.5-year period immediately subsequent to the launch of fluoxetine in the U.S. (1988-1991). The authors concluded that patients treated with fluoxetine had nearly 26% less depression-related health care expenditures than those treated with amitriptyline and over 28% less than those treated with nortriptyline. Again, no direct measure of effectiveness was included. However, patients initiating on amitriptyline were more than three times as likely to require a change in antidepressant pharmacotherapy, while patients initially prescribed nortriptyline were nearly four times more likely to change medication relative to patients initially prescribed fluoxetine. Similarly, 64% of patients initiating on fluoxetine received at least 180 days supply, whereas only 52% of patients on amitriptyline and 48% of patients on nortriptyline received at least this amount. The patients in this study were similar to those in the previous analysis.

         In the largest retrospective study conducted to date, Simon et al. compared overall mean total six-month health service costs for patients treated with fluoxetine (n=2,180) versus those for patients treated with imipramine (n=1,721) or desipramine (n=1,268).⁴⁶ This study, based on HMO medical claims, concluded that although treatment with fluoxetine increased total costs, the increase was not significant statistically. This study included no direct measure of treatment effectiveness.

         In a retrospective database study using 2-stage econometric modeling, Hylan et al. (1998) compared the costs of one year of treatment with fluoxetine (n=1,251) with those of treatment with TCAs (n=698).⁶⁸An intent-to-treat, multivariate analysis showed that patients who initiated therapy with fluoxetine had significantly lower total direct health care costs than patients initiating with TCAs. Treatment costs included costs of hospitalizations, ER, lab and physician visits, other outpatient services, and drugs. The patient claims included in this study were restricted to those for patients with a new episode of depression who were receiving initial therapy, selected from U.S. fee-for-service private insurance claims (MarketScan 1990 to 1994). These investigators controlled for observed and unobserved differences between treatment groups more explicitly than in the previous study by using 2-stage econometric modeling.

         Obenchain et al. (1997) conducted a retrospective bootstrap analysis to compare total annual charges incurred by patients taking fluoxetine, TCAs or trazadone for a new episode of depression.⁶⁹ On average, treatment with fluoxetine (n=799), relative to treatment with trazadone or TCAs (n=443), resulted in a negative incremental cost-effectiveness ratio (ICER). Specifically, fluoxetine realized a reduction in annual charges by $16.48 (0.2%) per patient for each percentage increase in effectiveness. However, the 95% upper confidence limit on the ICER was positive, implying that fluoxetine may actually increase annual per patient charges. The authors stated that any such increase was no more than $130 per patient (1.8%) for each percentage increase in the proportion of patients remaining stable on initial pharmacotherapy for at least 6 months.

         Data for this study were collected from the MarketScan database (1990-92) for a period of 12 months proceeding the index event of a diagnosis for depression and a prescription for antidepressant medication. A new episode was identified as a treatment-free period of four months prior to the index event. Patients participated in either fee-for-service or managed care health insurance plans for employed individuals and families. The principal measure of effectiveness was the proportion of patients remaining stable on the antidepressant for at least six consecutive months.

         While this study may provide a more reliable and widely applicable approach to statistical analyses of the cost-effectiveness of competing treatments, it has several limitations. For example, sufficient data were not available at the time of the study to make comparisons among SSRIs or to evaluate trazadone and TCAs in separate cohorts. Furthermore, the study relied on effectiveness measures from medical claims data, which do not provide detailed clinical outcomes, such as symptom scores.

      3. Paroxetine

         Several decision-analytic models have been developed which compare treatment with paroxetine to that with a TCA, namely imipramine. The models estimate that direct medical costs incurred with first-line treatment with paroxetine are no more--if not less--than costs for first-line treatment with imipramine. Several of these studies demonstrate that higher continuation rates (and lower dropout rates) of paroxetine contribute to its superior cost-effectiveness over TCAs. The models described here use results from randomized clinical trials to estimate treatment costs in the clinical practice setting. As a result, external validity of the studies is questionable. Comparative studies based in a naturalistic setting would be more useful in assessing the cost-effectiveness of paroxetine compared to TCAs.

         In 1995, Bentkover and Feighner designed a clinical decision analytic model to estimate the annual direct medical costs of treatment with paroxetine or imipramine. ⁷⁰ Efficacy and effectiveness probabilities were taken from a six-week clinical trial in 717 patients with major depression. The authors concluded that first-line treatment with paroxetine costs no more--if not less--than treatment with imipramine, due to reduced rates of relapse and fewer hospitalizations. Average annual direct medical costs were estimated to be $100 less for paroxetine than imipramine ($2,348 and $2,448, respectively). Sensitivity analysis reveals that the model is robust to changes in major parameters including hospitalization costs and relapse rates. However, it is sensitive to short-term dropout rates (total direct medical costs remain lower for paroxetine than for imipramine when the continuation rate is greater than 47% for paroxetine and constant at 46% for imipramine). It should be noted that this model potentially underestimates the total costs of treatment because it excludes indirect costs and treatment costs due to adverse events.

         The clinical trial data used in the above model were used to power two other decision-analytic models that estimate the costs of successful versus failed therapy with paroxetine and imipramine in the UK ⁷¹ and Canada.⁷² As with the Bentkover study, the models estimated treatment patterns from focus groups and obtained costs from health database claims (private and public for the UK and Canada, respectively). In the UK model, when the costs of successful treatment and dropouts were considered separately, successful treatment with paroxetine was associated with lower costs than with imipramine, indicating that poorer tolerance and higher dropout rate increased treatment costs. However, no difference in total costs between the two medications was found. One limitation of the UK study is that it assumes medication discontinuation rates differ by 20% between paroxetine and imipramine, a value greater than that used in other studies. The Canadian model demonstrated cost savings of $96 CAN per patient for the paroxetine group compared to imipramine.

         In all three models, sensitivity analysis revealed that continuation rates and dropout rates, rather than drug costs and relapse rates, are the most important predictors of the overall costs of care. For example, in the Canadian model, Paroxetine was cost effective only if the daily costs of medication were less than $4 CAN and the continuation rate was greater than 47%.⁷² Although the UK and Canadian models support the findings of the U.S. model, they are not necessarily generalizable to the US.

      4. Sertraline

         Studies comparing treatment with sertraline to that with TCAs are limited in methodology. Both of the studies reviewed below suggest that treatment with sertraline results in lower costs than treatment with TCAs. One study also asserts that sertraline is also more effective than TCAs. Taken together, it is reasonable to conclude that treatment with sertraline is more cost-effective than treatment with TCAs.

         Skaer et al. conducted a retrospective database analysis comparing direct health service expenditures for treatment with sertraline to that with TCAs (amitriptyline, desipramine and nortriptyline) in a U.S. managed care setting.⁷³ Multivariate regression analysis revealed that sertraline use resulted in significantly lower direct medical costs than TCAs for one year of treatment following initiation of therapy ($168 per patient, or 21% less than TCAs). Costs associated with depression-related physician visits and general and psychiatric hospitalizations were significantly lower for sertraline than for TCAs, while prescriptions costs were significantly higher and psychiatric costs were similar for sertraline and TCAs. The mean number of prescriptions (for a 30-day supply) was higher for the sertraline group than for TCA recipients (7.5 vs. 4.8) suggesting a more favorable tolerability profile with sertraline compared to TCAs. The study included 823 patients diagnosed with depression who were taking the study medications and for whom 18 months of health service data were available (6 months prior to and 12 months after receiving the initial prescription). Authors noted that this study is limited by its retrospective design, a lack of indirect cost data and the fact that HMO patients tend to be younger than the general population of patients with depression.

         Forder et al. used a retrospective quasi-experimental design to compare treatment costs and cost-effectiveness in patients who received sertraline (n=190) or TCAs (n=188) during a large clinical trial in the UK.⁷⁴ This model showed no significant differences in utilization of generic healthcare and social care services between the sertraline and TCA groups. However, mean costs associated with the use of inpatient and outpatient hospital care and community psychiatric nursing care were significantly lower in the sertraline group in both intent-to-treat and completer analyses. Overall mean costs were significantly lower for the sertraline group than for the TCA group as determined using some statistical tests, but not others. A significantly higher proportion of patients in both the intent-to-treat and completer analysis were rated by GPs as "somewhat improved" or "very much improved" for sertraline than for the TCA group. Resulting cost-effectiveness ratios favored sertraline for all definitions of costs and outcomes except for the acquisition costs of drugs alone.

         This study has been criticized for failing to include costs associated with work absence and/or reduced productivity into indirect costs and for the lack of an incremental cost-effectiveness analysis. In addition, opportunity costs, which are not easily validated, accounted for a rather high percentage of total costs. However, no significant differences were found between opportunity costs in the two groups.

      5. Venlafaxine

         Einarson et al. (1995) conducted a meta-analysis of randomized control trials to power a decision analytic model comparing the cost-effectiveness of treatment with venlafaxine to that of treatment with TCAs and trazadone.⁷⁵ Total inpatient depression-related medical costs were greater for patients treated with TCAs ($12,513), than for patients treated with venlafaxine ($12,201), or trazodone ($11,492). Total outpatient depression-related costs were again greatest for TCAs ($3,061), followed by venlafaxine ($2,401) and trazadone ($1,896). Medical care costs included costs of hospitalization, all physician visits, lab tests, radiology and medications. Venlafaxine patients experienced greater improvements in diagnostic test scores (62% reduction in HAM-D scores) than did patients treated with either TCAs (51%) or trazadone (49%).

         Cost-effectiveness ratios that calculated costs per symptom-free day in the inpatient setting were lowest for the venlafaxine group ($93), followed by trazadone ($108), and TCAs ($124). In the outpatient setting, trazadone was the most cost-effective agent at $20 per symptom free day, followed by venlafaxine ($23), and TCAs ($29). Although meta-analysis of randomized clinical trials is a powerful tool to obtain good probability estimates, the external validity of the RCT is questionable, especially when estimating outpatient costs.

         Using a similar framework as the U.S. model, Einarson et al. evaluated the cost-effectiveness of venlafaxine and TCAs from the perspective of the Canadian healthcare system.⁷⁶ In this analysis, venlafaxine was more cost-effective in terms of expected costs per success and per symptom free day (SFD) than TCAs both in the inpatient and outpatient setting. Venlafaxine was also dominant for all incremental pharmacoeconomic analyses. This model was robust for outpatients, but somewhat sensitive to changes in parameters for inpatients. In the worst case inpatient scenario, TCAs were superior to venlafaxine; in the worst case outpatient scenario, venlafaxine remained more cost-effective than TCAs.

         Griffiths et al. (1999) conducted a retrospective database analysis of nine U.S. managed healthcare plans to compare treatment with venlafaxine to treatment with TCAs⁷⁷. Patients were included who had switched from an SSRI after at least two consecutive months of therapy followed by at least two consecutive months of a TCA (n=172) or venlafaxine (n=188). Total direct healthcare costs were not significantly different between the two treatment groups. Further analysis by prescriber specialty and diagnosis code showed cost differences between the groups. While mean depression-coded costs were significantly greater for the venlafaxine group than for the TCA group ($1,948 vs. $1,396), mean non-depression coded costs were significantly lower. When psychiatrists were the initial prescribers of second-line therapy, significant cost differences were observed, but not with primary care physicians. In addition, patients receiving TCAs were more likely to visit an outpatient medical facility and a medical specialist other than a psychiatrist and were also more likely to receive anti-anxiety medications. The study suggests that the greater costs of pharmacotherapy with venlafaxine may be offset by a reduction in costs due to other medical services. The model is strengthened by considering observed and unobserved differences across patients. Limitations of this study include the lack of detailed clinical data on the severity of depression or the patients' complete history of antidepressant therapy. In addition, submitted charges were used as a proxy for costs. Overall the model suggests that venlafaxine is as cost-effective as TCAs for second-line therapy.

         The evidence on venlafaxine suggests that treatment with venlafaxine is more cost-effective than treatment with either TCAs or trazadone. Comparison of cost-effectiveness relative to SSRIs will be discussed later in the text.

      6. Nefazodone

         Four models compare nefazodone to TCAs. Two of these include fluoxetine in the comparison and are discussed later in this section. Although the studies discussed below differ on the relative costs of treatment with nefazodone, they do suggest that treatment with nefazodone is more cost-effective than treatment with imipramine.

         Using a Markov state-transition decision analytic model, Revicki et al. compared the lifetime costs and utility outcomes for three groups in a typical US managed care setting:⁷⁸

         • Treatment with nefazodone;
         • Treatment with imipramine; and
         • Step care with imipramine followed by nefazodone.

         Results of the base case analysis estimated lifetime medical costs (discounted at 5%) to be lower for imipramine and step care than for nefazodone ($15,348, $16,061, and $16,669, respectively). However, QALYs were greater for nefazodone treatment than for imipramine or the step approach (14.64, 14.32, and 14.40, respectively). The resulting cost-effectiveness ratio comparing nefazodone with imipramine was $4,065/QALY. This cost falls within the suggested criteria to be adopted into the Canadian healthcare system and may be equally, if not more acceptable in the U.S.

         The model was robust to adjustments in key parameters such as compliance rates, discount rates and maintenance treatment probabilities. Sensitivity analyses demonstrated that cost-effectiveness ratios for nefazodone compared with imipramine ranged from $2,572 to $5,841 per QALY gained. Comparing nefazodone to the step approach, cost-effectiveness ratios ranged from $1,436 to $6,764 per QALY gained. Limitations of the study include the omission of indirect costs, the assumption that the target patient population was 30-year old women, and the general limitations of economic models which rely on clinical trial data and expert opinion. The study is also limited in its applicability to the managed care setting. Overall, the study demonstrates that nefazodone is cost-effective compared to imipramine or step care with imipramine and nefazodone.

         Unlike the U.S. model, an earlier Canadian model estimates medical costs for nefazodone to be lower than those for imipramine. These differences are explained by differences in model parameters such as hospitalization, compliance rates and depression recurrence rates.⁷⁸

   2. Newer Antidepressants Compared

      Although most researchers have focussed on the cost-effectiveness of newer antidepressants compared to their older competitors, a few studies have been conducted comparing the cost-effectiveness of two or more new-generation antidepressants. Results have varied in these studies, and there is no apparent consensus that any one new agent is superior either in terms of effectiveness or cost-effectiveness.

      1. Fluoxetine Compared to Sertraline

         Boyer et al. performed the only prospective, double-blind, randomized, trial to date which compares treatment costs within the class of SSRIs.⁷⁹ Conducted in a primary care setting in France, this six month multicenter study compared treatment with fluoxetine (n=120) and sertraline (n=122) in patients with major depression. In this study, treatment with fluoxetine was associated with greater treatment costs than was treatment with sertraline, while clinical outcomes appear similar. Clinical outcomes revealed equivalent efficacy for sertraline and fluoxetine at mean dosages of 55 mg and 26 mg, respectively. 65% and 67% of patients remained at their starting dosage and mean duration of therapy was 153 and 158 days for sertraline and fluoxetine, respectively. Total costs per patient from the insurer's perspective were FF2936 for sertraline and FF3224 for fluoxetine; societal costs were FF7780 and FF8706, respectively. Per patient savings in medical resources utilization were FF23 less for sertraline than fluoxetine from the insurer perspective and FF74 less from the societal perspective. Per patient savings due to depression-related absenteeism were FF118 less for sertraline than fluoxetine from the insurer perspective and FF443 less from the societal perspective. Statistical data were not provided.

         Most of the total costs incurred in both groups (67.0% of total costs for the sertraline group and 69.7% of the cost for the fluoxetine groups) were due to indirect costs, primarily due to absences from work. Overall, the difference in total costs between the two groups was mostly due to differences in indirect costs.

         For this study, clinical and quality of life assessments were made at study entry and after four and six months of treatment. Patients with at least a 50% decrease in MADRS scores at 4 months that was maintained for 2 additional months were considered responders. Direct and indirect costs were calculated from both the societal perspective (patient out-of-pocket costs plus sickness insurance reimbursement) and the healthcare provider (sickness insurance) perspective. Total costs were assessed as the sum of direct and indirect costs related to medication, all physician visits, medical investigations, hospitalizations, paramedical procedures, work absenteeism, and lost productivity. Sickness insurance costs were determined by mean reimbursement rates by a French national health insurance fund for salaried workers. Societal costs were calculated on the basis of market prices or price rates for health products and services. Because sertraline was not available on the French market at the time of the study, no costs were attributed to either of the study medications for both direct and indirect cost analyses.

         Despite the difficulty of extrapolating these results to the U.S. healthcare system, the prospective trial design and naturalistic setting make this study more reliable than previous analyses. As such, the results of this study suggest that sertraline is more cost-effective than fluoxetine, although these results should be interpreted with caution.

      2. Fluoxetine Compared to Sertraline and Paroxetine

         Both of the following studies evaluate the cost of treatment of depression with fluoxetine versus those same costs of treatment with sertraline and paroxetine. These studies, based on retrospective analysis of claims data with or without econometric modeling show fluoxetine to be superior to the other SSRIs in terms of cost of treatment. However, these studies use data collected shortly after the launch date of the newer SSRIs. Outcomes may be skewed due to physician unfamiliarity with the drug and the fact that paroxetine and sertraline may have been used early on as second-line therapy for patients who discontinued fluoxetine.

         In a retrospective database study described above, Hylan et al. also compared the costs of one year of treatment with fluoxetine (n=1,251) to those for paroxetine (n=136) and sertraline (n=508).⁶⁸ Univariate analysis found no significant differences in mean health care costs across the three antidepressant medications. However, significant differences (p<0.0001) were found in median total and mental health care costs. Total median healthcare costs were $3,466, $3,491 and $3,613 per patient for fluoxetine, sertraline and paroxetine, respectively. Median mental health costs were $1,187, $970, and $1,048 per patient, respectively.

         Multivariate analysis, controlling for observed and unobserved characteristics, found that only patients who initiated therapy with fluoxetine had significantly lower total direct health care costs than with TCAs. Patients initiating therapy on sertraline or paroxetine did not incur lower total direct health care costs than did patients initiating with TCAs. In addition, patients initially treated with fluoxetine had significantly lower total health care costs than those treated with sertraline. This result contradicts that found in the Boyer study. Hylan et al. found no difference in total health care costs between the fluoxetine and the paroxetine groups. Total mental health care costs were not significantly different across the three medications.

         These findings should be interpreted cautiously for several reasons. First it should be noted that a dosage of 50mg/day was used for sertraline instead of the later recommended dosage of 100mg/day, which could have impacted clinical outcomes. More importantly, the analyses were conducted using claims from 1990-1994, a period when paroxetine and sertraline were recently introduced to market. Therefore, prescribers may have been more familiar with fluoxetine than paroxetine or sertraline and may not have used the newer SSRIs in the most efficacious and economical way. As the authors suggest, it would be useful to replicate this analysis using more current data, following patients over a longer time span.

         In 1995 Sclar et al. performed a similar analysis using insurance claims data from a network model HMO system.⁸⁰ Multivariate regression analysis estimated that total direct depression-related health service expenditures per patient were less for fluoxetine than for paroxetine or sertraline ($442,29, $744.06, and $756.23, respectively).The percentage of patients requiring dosage titration between the first and last prescriptions were greatest for sertraline (40.3%), followed by paroxetine (28.1%) and fluoxetine (16.1%).

         Again, these results should be interpreted with caution for several reasons. As with the previously described analysis by Hylan et al., paroxetine and sertraline had been newly introduced during the study period. Therefore, increased treatment costs may reflect physician practices that could have improved since the time of the study. The managed care patient population was homogenous and potentially younger and with fewer comorbidities than a general population of patients with depression. Furthermore, initial dosages and sources of cost data are not clearly defined. Finally, unlike the study by Hylan, unobserved differences across patients were not considered and may have skewed results.

         Given the results of the Boyer study in the previous section, it is reasonable to conclude that no clear distinctions exist between these SSRIs--fluoxetine, sertraline and paroxetine--with respect to either costs or cost-effectiveness of treatment.

      3. Venlafaxine Compared to SSRIs

         The studies by Einarson et al. discussed above also included a comparison of venlafaxine to SSRIs.⁷⁵ The results of these studies exhibit mild evidence for venlafaxine being more cost-effective than SSRIs. From the U.S. healthcare payer perspective, total inpatient depression-related costs were greater for venlafaxine ($12,201) than for SSRIs ($11,864), while total outpatient depression-related costs were greater for SSRIs ($2,412) than for venlafaxine ($2,401). Venlafaxine patients experienced greater reductions in their HAM-D scores (62%) than did patients on SSRIs (30%). Cost-effectiveness ratios calculated that costs per symptom-free day in the inpatient setting were lower for venlafaxine ($93) than SSRIs ($146). This superiority was maintained in the outpatient setting: venlafaxine ($23), SSRIs ($25).

         In the Canadian application of this model, these researchers found that venlafaxine was more cost-effective in terms of expected costs per success and per symptom free day than were the SSRIs in both the inpatient and the outpatient settings. Venlafaxine was also dominant for all incremental pharmacoeconomic analyses. However, in sensitivity analyses, SSRIs were more cost-effective than was venlafaxine in an inpatient worst case scenario, although venlafaxine remained dominant in all outpatient analyses.⁷⁶

C. The Cost-Effectiveness of Newer Antipsychotic Medications

1. The Cost of Schizophrenia and Related Illness in the United States

   Cost of illness studies on psychotic disorders in the U.S. reveal that patients with schizophrenia consume a disproportionate percentage of national healthcare dollars. It has been reported that although schizophrenia occurs in only 1% of the general population, patients with the disorder consume approximately 2.5% of total U.S. healthcare dollars and 22% of the costs of all mental illness.^50,51,81 Furthermore, patients with schizophrenia consume 10% of Social Security benefits, occupy up to 25% of all hospital beds, and account for 40% of long-term care days.⁵¹

   Rupp & Keith estimated the total cost of schizophrenia in 1990 to be $33 billion.⁸¹ Of this amount, $17 billion was attributable to direct medical costs while $12 billion were attributable to indirect costs.⁸² Wyatt et al. estimated the indirect cost of schizophrenia to be much higher by using a "productive person-years" method.⁸³ These researchers estimated the total costs of schizophrenia in the U.S. in 1991 to be $65 billion ($19 billion in direct costs and $46 billion in indirect costs).⁸² It is important to note that these cost of illness studies were conducted before the introduction of many of the antipsychotic agents discussed in this report. Since the early 1990s, atypical agents such as risperidone, clozapine and olanzapine may have contributed to a reduction in total costs or the proportion of U.S. health care dollars consumed by patients with schizophrenia.

2. Pharmacoeconomic Analyses of Specific Antipsychotic Agents

   We reviewed 38 studies that evaluate the pharmacoeconomics of newer antipsychotic medications to a relevant level of rigor. Only a small number of studies are prospective, randomized economic clinical trials conducted either in a naturalistic health care setting or as a "piggy back" to a clinical trial. Many of these studies are retrospective analyses of claims data from public and private health care payers. Several other studies are decision-analytic models that incorporate both claims data, the results of randomized clinical efficacy trials, and the results of expert panel interviews.

   Most studies conducted to date in schizophrenia are not technically "cost-effectiveness" studies that measure cost per successful patient outcome over time, but cost-minimization analyses that measure total cost savings per patient over time. An ideal study would explicitly measure direct and indirect medical costs associated with the use of newer antipsychotic medications. Until recently, most published studies have used only a proxy for these costs (such as reduction in hospital days) to estimate cost-effectiveness.⁵⁴ This proxy is useful because traditionally a majority of medical costs incurred in treating patients with schizophrenia are due to hospital costs. However, full enumeration of total medical costs is desirable.

   As for antidepressants, many of these studies have been sponsored by pharmaceutical manufacturers. However, 12 of these studies were funded at least in part by government agencies or private foundations. Nonetheless, one might expect publication bias to color the results. In fact, in all the studies reviewed below, treatment with the agent manufactured by the sponsor appears to be at least no more costly in terms of total direct medical costs than treatment with the comparative agents. The claim that one newer agent is economically superior over another may appear inherently specious if the study is sponsored by the manufacturer claiming superiority. However, the result that treatment with new-generation antipsychotics is at worst no more costly in terms total direct medical costs than treatment with older agents is confirmed by studies sponsored by government agencies or other third parties.

   The following discussion is limited to studies conducted from the perspective of the US healthcare system. Reference is made to studies conducted in other countries when they provide unique data or the results compare and contrast to those for the US.

   1. Studies of Single Agents that Measure Impact on Hospital Admissions and/or Stays

      We reviewed several historical comparator studies for risperidone which report data on reduction in hospital days and/or number of admissions one or two years after treatment compared with one year prior to treatment. Most of these studies do not report cost outcomes. Those that do will be discussed later in this report. Results from intent-to-treat analyses were variable. Treatment with risperidone resulted in anywhere from a 1% increase in hospitalization days to a 58% reduction in hospitalization days compared to the time period before treatment with risperidone ^82,84-88 Completer (i.e., patients who did not drop out of the study) analyses consistently demonstrated improvement with risperidone treatment, which resulted in anywhere from a 20% to 74% decrease in hospitalization days.^{84,85,87,89,90} Where reported, the number of admissions either remained constant ⁸⁴ or were reduced (up to 67%).^82,86-88 (See Exhibit VII-2.)

      Several clozapine studies also use hospitalization as a proxy for cost.^91,92 These studies indicate a reduction in hospital days over a 2.5 year period and some improvement in patients who remain hospitalized. (See Exhibit VII-2.)

      These types studies have several limitations. (The reader is referred to Foster and Goa 1998 for a more detailed explanation.⁵⁴) First, the retrospective historical comparator design may lead to historical bias. In the absence of a control group, it is difficult to determine the extent to which improvement in the post-index period is attributable to the study medication. Most of the studies do not specify which type of treatment (e.g., pharmacotherapy, counseling) patients were receiving prior to treatment with risperidone. Furthermore, in their review of risperidone, Foster and Goa note that with the movement away from inpatient to community care, treatment patterns may have changed during the study period.⁵⁴ Although two studies reviewed concluded that changes in policy did not explain the reduction in hospital days by risperidone.^82,93 In addition, inconsistencies in the criteria for hospital admission, patient demographics, the availability of hospital beds and other external factors can affect hospitalization rates. Therefore hospitalization data may not be an ideal measure of cost-effectiveness. Finally, long-term studies indicate that healthcare utilization in patients with schizophrenia declines over time due to several factors.⁵⁴ Therefore, real time studies would provide more reliable data than would historical comparisons.

      Most of the studies reviewed concerning olanzapine, quetiapine and clozapine which report hospitalization rates incorporate these data into overall medical costs. Hospitalization data for these treatments will not be reviewed independently, but will be incorporated into more detailed reviews of these studies.

2. Comparisons Among Antipsychotics
   1. Newer Antipsychotics Compared to Older Antipsychotics

      Several economic models suggest that, despite their higher medication costs, risperidone, olanzapine, and clozapine may be more cost-effective than older antipsychotics (e.g., haloperidol) in the treatment of schizophrenia.

      Palmer, et al constructed a five-year Markov model to compare the cost of treatment of schizophrenia in the U.S. using olanzapine, haloperidol, or risperidone.⁹⁴ Results were normalized to 1995 U.S. dollars. For the five-year period of the model, patients treated with olanzapine experienced 6-7 months more time in a disability-free state (Brief Psychiatric Rating Scale (BPRS) score <18) than those treated with haloperidol. The total direct medical costs of treating patients with olanzapine over the five-year period were $1,539 less than for haloperidol and $1,875 less than for risperidone. Patients treated with olanzapine first-line showed a 1.6% reduction in total costs versus haloperidol. Cost savings were due to reduced hospitalization rates, relapse rates and lower suicide rates.

      The analysis was robust for variations in the discount rate, the rate and/or costs of suicide attempts, frequency duration and/or cost of hospitalization and/or other medical resource utilization estimates. However, the model was sensitive to drug dosages of olanzapine and risperidone and a decrease in length of stay. Risperidone becomes more cost-effective than olanzapine when the dosages are lowered and raised, respectively. Olanzapine is no longer cost saving when the initial duration of hospitalization is halved. As a result, the model has been criticized for using a dose of olanzapine (10 mg/day) that is lower than generally used in clinical practice, while using a dose of risperidone (6 mg/day) that is much higher than generally used. Furthermore, the high risperidone dosages may aggravate side effects (e.g., extrapyramidal syndrome (EPS)) and limit the real world effectiveness of risperidone in the simulation.

      This model was populated using probability estimates obtained from a one-year international comparative trial of olanzapine (10mg/day) versus haloperidol (15mg/day) and a 28-week comparative trial of olanzapine (10mg/day) versus risperidone (6mg/day).^41,95 Additional inputs to this model were obtained from the medical literature and an international advisory panel of psychiatrists and health economists. This expert opinion was used to extrapolate clinical data from 1 year or 28 weeks to the 5-year timeline--one of the major limitations of this model. The model examined the cost of treating patients experiencing multiple schizophrenia episodes, while excluding treatment-resistant patients or those experiencing a first-break episode.

      A recent historical comparator pilot study by Galvin examined the medical records of 37 patients with schizophrenia, schizoaffective disorder or bipolar disorder one year before and one year after treatment with atypical agents (risperidone and clozapine).⁹⁶ Based upon patients' detailed medical records, the study estimates an annual total cost savings of $3000 per patient in the community mental health setting when treatment is moved to an atypical agent. There are several limitations inherent to this small-scale community-based study. For example, the entry criteria may have introduced a selection bias by excluding clients who were most intolerant or least responsive to older antipsychotic medications and could not continue their use for one year. Although the study was small, uncontrolled, and retrospective in design, its focus on the community mental health setting is more naturalistic than large-scale clinical trials and set an important example for future research.

   2. Risperidone Compared to Haloperidol

      The pharmacoeconomic evidence of risperidone versus typical antipsychotics indicates that the costs incurred for treatment with risperidone approximately equal those incurred for treatment with haloperidol. Results from these studies suggest that risperidone, despite its higher acquisition cost, does not significantly increase overall treatment costs compared to conventional antipsychotics and is at least as effective. However, several considerations raised by Zito suggest that experimental design and reporting of data from the large multi-center trials of risperidone compared to haloperidol favored clinical and economic outcomes for the atypical agent.⁵⁰ On the other hand, economic analyses that assume equal compliance rates or effectiveness for risperidone and haloperidol may underestimate potential cost savings with risperidone.^97,98

      Janssen Pharmaceuticals, the manufacturer of risperidone, is currently sponsoring a multicenter, randomized clinical trial across 21 investigative sites in the U.S. The trial, named the Risperidone Outcomes Study of Effectiveness (ROSE), compares treatment with risperidone to 13 conventional agents in patients with schizophrenia or schizoaffective disorder. The trial examines clinical, quality of life, and economic outcomes in a typical U.S. clinical practice setting. Clinical outcomes include changes in psychiatric symptoms, side effects, health-related quality of life, drug satisfaction, therapy switching, rehospitalization for relapse and use of psychiatric services. Economic outcomes focus on the costs of mental health-specific care. ⁹⁹

      In 1997, Mahmoud et al. presented a cost analysis of the ROSE trial at the 36^th Annual Meeting of the American College of Neuropsychopharmacology. Estimates of national average costs were used to calculate mean costs per patient. The intent-to-treat analysis favored conventional agents over risperidone in total, acute care, outpatient and medication costs. Intent-to-treat analyses included patients having prolonged drug-free periods, polypharmacy, or crossing over to receive alternative treatment; patients who remained in the treatment arm (50% of those randomized to risperidone and 61% of those randomized to conventional agents) may have received multiple or concurrent antipsychotic agents.¹⁰⁰ Total costs for treatment with risperidone were $1,963 higher than for treatment with conventional agents, but the results were not judged to be statistically significant. When patients switching therapies were excluded from analysis yields, risperidone was cost saving compared to conventional agents in total, acute care and outpatient treatment costs. Although the total costs differed between the types of analysis, mean positive and negative scores on the PANSS showed greater improvement at one year with risperidone than with conventional agents in both types of analyses. Further publication of trial data was not found at the time of this study.

      Davies et al. developed a robust decision-analytic model to compare treatment with risperidone or haloperidol over a two-year period in patients with chronic schizophrenia.¹⁰¹ The authors constructed a decision tree that considered ten possible clinical scenarios. The expected cost per favorable outcome was $11,395 (1993 Australian dollars) less for risperidone than haloperidol over the two year period (cost per patient treated was $2,783 less for risperidone than haloperidol). Direct costs included cost of drugs, general practitioner visits, psychiatrist visits, hospitalization, social worker consultations, lab tests and hotel accommodation. Probability estimates (e.g., efficacy, tolerability, and drop-out rate) were derived from meta-analysis of clinical trials and expert opinion. The model was robust to sensitivity analysis and the break-even point for treatments to be of equal cost was an 11.8% difference in effectiveness. Using baseline assumptions, risperidone was found to be 20% more effective than haloperidol. It should be noted that costs, reported in Australian dollars, may not necessarily be applicable to the U.S. setting.

      Glazer & Ereshefsky developed a US-based decision-analytic model which is less robust than the one described above.⁹⁸ The model estimated which of three possible treatment paths would have the lowest total direct costs during the first year after discharge from a hospital for a 29-year old male with a history of relapse and rehospitalization for schizophrenia. While compliance rates were varied for each of the three treatment options (65% for risperidone 6mg/day, 50% for oral haloperidol 15mg/day, and 80% for depot haloperidol decanoate, 150 mg/month), clinical outcomes probabilities were assumed the same for each treatment. Total direct treatment costs were less for depot haloperidol decanoate than for continuing oral haloperidol, and both were less than for treatment with risperidone. Sensitivity analysis showed that when the compliance rates for risperidone were increased to 80% and assuming 25% discount on the cost of medication, total one-year direct costs with risperidone were $1,862 less than with oral haloperidol and $705 less than haloperidol decanoate.

      We have reviewed 11 historical comparator studies which examine the costs of treatment one year before and one to two years after treatment with risperidone. The pretreatment costs may have included treatment with haloperidol or other typical antipscyhotic agents. Costs were obtained from regional medical claims databases or unspecified sources. In most cases (5 of 7 ITT analyses^{82,10284-86,88,103} and 4 of 4 completer analyses^{84,89,93,104,105}) risperidone resulted in overall cost savings (range, $308 to $2596 in U.S., 1993-1997 dollars). The historical comparator studies that used intent-to-treat analyses (i.e., including cross-over and treatment-resistant patients, etc.) found that the reduction in costs with risperidone was generally smaller than that observed in studies that included only treatment successes. Furthermore, these studies generally find that savings are more significant in the second year of therapy (cost savings may approach 50%), suggesting that long-term comparative economic studies are needed to demonstrate the true cost-effectiveness of antipsychotic treatment.

      Only two of the historical comparator studies found that treatment with risperidone increased total costs over treatment with haloperidol when crossover patients were included.^84,88 However, in neither of these cases, did treatment costs increase by more than 3%. That one of these studies found inpatient costs to be 10% higher after treatment is misleading.⁵⁴ In this study, inpatient costs included all inpatient care in hospitals, skilled nursing facilities, psychiatric facilities and department of correction facilities. After initiation of risperidone, acute inpatient care costs actually decreased, while other inpatient costs increased. Furthermore, in both of these studies, there was a very low rate of acute inpatient care before initiation of treatment. Therefore, substantial cost savings could not be achieved.⁵⁴

      A more recent retrospective study conducted by Schiller et al. compared treatment with risperidone to conventional therapies in a community mental health setting.¹⁰⁶ This study used a quasiexperimental mixed-model analysis and obtained outcomes and service utilization data from outpatient charts and databases. Results found no difference in clinical outcomes, measured by Global Assessment of Functioning (GAF) scores, but a trend towards higher total costs for risperidone. The $370 increase in total costs with risperidone was attributed mainly to higher medication costs and medication-related outpatient visits, but only the difference in medication costs was statistically significant. Mental health service costs and lab costs were obtained from county year-end (1994) cost reports of San Francisco community mental health services. Medication costs were average wholesale prices. One limitation of this study is the comparability of the two groups. The risperidone group was selected based upon a change in medication, whereas the comparison group was not. The authors note that a higher service utilization in the risperidone group in the pre-risperidone period may have produced possible selection bias. The authors suggest that given the possible selection bias, restricting use of risperidone based upon an increase in costs is not justified.

      Nightengale (1995) conducted a small naturalistic retrospective cohort study which estimated the monthly costs for risperidone (n=28) compared to haloperidol (n=24) inpatients with schizophrenia, schizoaffective disorder or major depression with psychotic features. ⁹⁷ The study found no significant differences in length of follow-up, length of stay, or hospitalization rates between the two treatment groups. The haloperidol cohort used significantly fewer physician services than did the risperidone cohort (every 127 days versus every 75 days). However, haloperidol patients were more likely to be admitted to day hospitals. Although the difference in monthly costs per patient between the risperidone and haloperidol groups were not judged by the authors to be significant ($1,636 versus $1,759, p=0.04693), the total monthly cost per patient was considerably lower in the risperidone cohort ($123.24 less per patient than haloperidol). Although the setting of care is realistic, the 100% compliance rate assumed by the authors for both medications is not. Considering that compliance with risperidone may be greater than with haloperidol in reality, the study may have underestimated cost savings with risperidone.

   3. Olanzapine Compared to Haloperidol

      Studies comparing olanzapine to haloperidol provide solid evidence that olanzapine is more cost-effective than haloperidol.

      Several prospective economic analyses have been conducted as "piggy back" studies to an international comparative trial of olanzapine versus haloperidol.⁹⁵ This trial enrolled patients with schizophrenia, schizophreniform disorder or schizoaffective disorder were followed during a 6-week acute phase (n=551, olanzapine; n=266, haloperidol) and 46-week maintenance phase (n=270, olanzapine; n=74, haloperidol). Clinical and resource utilization data were collected in 174 clinical centers in 17 countries, 66 of which were in the US. Although prospective studies such as these are ideal, the analyses were limited because economic data were not collected for patients after they withdrew. Several economic models imputed missing values to determine incremental cost-effectiveness. Several of the analyses are published in full^94,107-110, while others have only appeared as abstracts or presentations (Dutch and German analyses)^111,112 Overall, the prospective analyses indicate that olanzapine is associated with lower total costs than haloperidol and possibly risperidone in moderately to severely ill patients with chronic schizophrenia.

      In a basic cost analysis, Hamilton et al. used resource utilization data from an international trial to estimate direct medical costs for those patients who completed treatment with olanzapine or haloperidol.¹⁰⁸ Olanzapine showed a trend toward greater efficacy compared to haloperidol in the acute phase based upon the BPRS scale. This trend did not attain statistical significance in the maintenance phase according to BPRS, PANSS and QLS scores. Total costs were significantly lower for olanzapine in the acute phase ($388 lower over 6 weeks) but not in the maintenance phase ($636 lower over 46 weeks). These results are considered conservative because patients who discontinued therapy were not considered. It should be noted that significantly fewer patients treated with haloperidol completed the acute phase of treatment than did patients treated with olanzapine (43% versus 64%).

      Subsequently, Obenchain et al. developed a mixed effects linear model to estimate cost-effectiveness of olanzapine compared to haloperidol over the one-year study period of the international trial. ¹⁰⁷ The model used homogenous variance components and imputed missing values for patients who discontinued treatment early. Total direct medical expenditures per patient were estimated to be $10,301 lower for patients treated with olanzapine than haloperidol (p<0.0001). Patients taking olanzapine experienced on average 18.3 more symptom-free days than those taking haloperidol (symptom-free = BPRS score <18 on items 0 to 6). The incremental cost-effectiveness ratio estimated annual savings of $563 per symptom-free day gained. Mean annual hospitalization costs were $10,856 less with olanzapine (p=0.007). Bootstrap analysis was highly stable for the finding that olanzapine is associated with greater efficacy and lower costs. However, results were somewhat sensitive to drug costs, which were based upon dosages of 10mg/day for olanzapine and 15mg/day for haloperidol. If recalculated on the basis of the cost of olanzapine at 20 mg/day, the annual savings are reduced to $7,542 per patient.

      Two retrospective historical comparator studies compared costs for six months before and after treatment with olanzapine.^113,114 Results of both studies showed that total medical costs did not increase (and may have decreased) despite an increase in medication costs. As discussed earlier, these studies have significant limitations including potential selection artifacts and historical bias. ⁵³

   4. Quetiapine Compared to Usual Care

      In light of the previously published, and often poorly designed, cost-effectiveness literature on risperidone and olanzapine, the manufacturers of quetiapine are conducting a well-designed naturalistic comparative pharmacoeconomic trial. Hong et al. have described the design of a multicenter, open-label, randomized, comparative, effectiveness trial for treatment with quetiapine or "usual care" in revolving door patients.¹¹⁵. The primary objective of the trial is to compare the rates of rehospitalization due to psychiatric, medical, or social consequences of schizophrenia or schizoaffective disorder for revolving door patients treated with quetiapine or usual care. In this study, "usual care" is defined as treatment with any first-line oral or depot antipsychotic -- including atypicals, excluding clozapine -- commercially available at the time of the trial in the United States. (Clozapine is excluded from the study because it is not a first-line antipsychotic and clozapine-treated cohorts are biased towards treatment-resistance and chronicity.) The study will also assess the use of inpatient and outpatient health services, quality of life (SF-36, BASIS-32), and outcomes such as efficacy, safety, compliance, and patient satisfaction. The trial duration is 53 weeks.

      The study design for this trial focused on a subpopulation of patients with the greatest potential for cost savings by treatment with quetiapine or usual care. The appropriate minimal threshold for revolving-door status was calculated from hospitalization rates and inpatient payments obtained from MedStat MarketScan databases (1/91-12/92) and Medi-Cal data (7/91-6/93). Analysis of medical claims databases indicated that patients with schizophrenia with an annual hospitalization rate of at least 1.0 (=10% of total sample) generate about 50% of inpatient costs. Therefore the minimal threshold for revolving-door status was defined as 1.0 admission per year. It was estimated that 182 patients would be necessary for each arm of the trial to detect a 15% difference in rehospitalization rates, the value expected to result in significant reduction in hospitalization costs.

      Preliminary assessments indicate patients had had two hospitalizations in the past two years, one hospitalization and one admission to a crisis shelter for conditions related to schizophrenia or schizoaffective disorder in the past year, or at least three admissions to a crisis shelter in the past year. Settings included U.S. clinical settings associated with community mental health centers, state psychiatric facilities, private psychiatric hospitals, university hospitals, and VA hospitals. No further publication of these trial results was found at the time of this study.

   5. Risperidone Compared to Olanzapine

      Current inconsistencies regarding the dosages in the risperidone Compared to olanzapine studies render the evidence inconclusive as to which agent is more cost-effective.

      Grainger et al. presented preliminary results of a resource utilization and quality of life analysis of the 28-week clinical trial comparing risperidone and olanzapine.¹¹⁶ In the 91 evaluated patients, mean total direct medical costs were $493 lower per patient with olanzapine than risperidone (p<0.01). This figure was statistically significant after controlling for hospitalization status at baseline and length of hospitalization prior to randomization. Mean medication costs were significantly higher for olanzapine than risperidone (+$655). However, mean inpatient and outpatient costs per patient were significantly lower for olanzapine (-$318 and -$829, respectively). These preliminary results must be interpreted with caution, because of the small sample size and because the study is not yet published in full. Furthermore, costs for nonresponders and discontinuers were not assessed beyond the acute treatment phase. Finally, the study has been criticized because it uses doses for risperidone that are higher (6 mg/day) and doses of olanzapine that are lower (10 mg/day) than is customary in clinical practice.⁵³

      A more recent short-term comparative study used more realistic dosages for risperidone and olanzapine (4.8 and 12.5 mg/day, respectively). Risperidone was potentially more effective and as safe as olanzapine and it was suggested that treatment costs would be lower with risperidone. ¹¹⁷

   6. Clozapine Compared to Typical Antipsychotics for Treatment Refractory Schizophrenia

      Collectively, the results of the numerous following studies together provide strong evidence of clozapine's superior cost-effectiveness relative to older, typical antipsychotics for treatment-resistant schizophrenia.

      Clozapine, the first atypical antipsychotic, has been studied extensively for treatment-resistant schizophrenia. The side effects common to traditional antipsychotics are not found with clozapine (EPS, tardive dyskinesia, etc.). However, the risk of agranulocytosis is greater with clozapine and it is reserved for treatment-refractory cases. Generally the higher acquisition cost with clozapine is offset by a reduction in hospital days. Overall, studies demonstrate that clozapine is no more costly, and more effective than conventional therapy in patients with treatment-resistant schizophrenia.

      Rosenheck et al. completed the only prospective, randomized double-blind comparative study evaluating the cost-effectiveness of clozapine (n=205, 552mg/day) versus haloperidol (n=218, 28mg/day).¹¹⁸ Patients were older males with treatment refractory schizophrenia and a history of high inpatient service utilization (30-364 days in previous year). An intent-to-treat analysis showed no significant cost savings associated with treatment with clozapine (total health costs were $2,441 less and societal costs were $2,773 less for clozapine than haloperidol). Outpatient treatment and medication costs were substantially more costly for the clozapine group than for the haloperidol group. These costs were offset by fewer hospital days in the clozapine group (24.3 fewer hospital days). Treatment with clozapine was also associated with significantly better clinical outcomes. The clozapine group showed improvements in PANSS scores relative to the haloperidol group, as well as significantly higher scores on QOL and EPS assessment scales. Overall, clozapine was cost neutral, but more clinically effective than haloperidol. This study was considered to be a more definitive cost-effectiveness analysis than previous studies.

      Essock et al. reported results from a prospective randomized open-label study comparing clozapine (n=138, 486mg/day) with standard care (n=89, neuroleptic/psychosocial services).¹¹⁹ Patients were screened in 1991 from Connecticut's State psychiatric hospitals for treatment-resistant schizophrenia (>=2 drug trials) or schizoaffective disorder. The trial lasted two years. Hospitalization (i.e., time to readmission and probability of discharge) was used as a proxy for costs. Preliminary data show decrease in the rate of readmission for those treated with clozapine compared to those treated with typical agents (3% versus 29% at 6 months and 17 versus 41% at 1 year for clozapine versus typical agents). However, the probability of discharge was not significantly better for clozapine than for standard care. BPRS and QOL scores reveal that patients treated with clozapine show significantly reduced side effects and disruptiveness than those treated with conventional antipsychotics. However, clozapine was no more effective than were typical agents in reducing symptoms, improving QOL or probability of life.

      Another historical comparator retrospective study compares clozapine (n=133) to conventionals (n=51) one year prior to and two years after therapy in treatment-resistant patients.¹²⁰ Total costs for the clozapine group were increased in the first year after treatment but decreased in the second year of treatment for both ITT and completer analysis. Cost-reductions were more substantial in completer analysis, but ignoring the cost of drop-outs is misleading.⁵⁰ The net treatment costs over the three-year period were $1,029 greater for the clozapine group than the neuroleptic group for those patients who completed treatment (for clozapine, n=87 one year before and first year after index date; n=46 in second year). However, assuming a shorter initial hospital stay (14 days), clozapine was found to be more cost-saving (3-year net savings of $15,036).

      Another historical comparator study compares the cost of treatment with clozapine (n=86) to conventional therapy (n not stated) five years prior to and one and two years after treatment. Costs and probability estimates were taken from manufacturers' quarterly reports and patient records from a clinical investigation.¹²¹ Clozapine patients were participating in a clinical trial in the inpatient setting. Patients receiving conventional therapies were considerably less seriously ill than clozapine patients (which may underestimate the benefit of clozapine). The study found an average cost savings of $20,000 by the second year for patients who completed two years of treatment with clozapine versus patients who were treated with conventional antipsychotics. However, the study did not measure clinical outcomes for the control group. The rehospitalization rate was substantially lower for clozapine versus conventional antipsychotics (27.8% versus 56.2%). Total hospital costs from the five-year period before treatment to one and two years after treatment went from $73,403 to $58,941 to $28,367 with clozapine and $72,121 to $67,934 to $61,518 with conventional therapy, demonstrating substantial cost-savings in the second year.

      A small retrospective historical comparator study evaluated clozapine responders with two years of follow-up (n=37) in a university-based ambulatory clinic. Cost data (inpatient and outpatient treatment, housing costs, other costs, and family burden) were available for 47 of the 96 treatment-resistant patients who participated in a clinical study. Outcome measurements were psychopathology, quality of life, global functioning, work function and rehospitalization. Results indicated that a dramatic decrease in the frequency and cost of rehospitalization contributed to a significant decrease in cost of treatment for patients who continued clozapine treatment for at least two years. For the 37 patients who continued clozapine treatment for two years, costs decreased by $22,936 per year per patient. Including drop-outs, there was a savings of $8,702/year/patient. The study is limited by the retrospective collection of cost data, on average three years after the start of treatment with clozapine.¹²²

      The results of these numerous studies together provide strong evidence of clozapine's superior cost-effectiveness relative to older, typical antipsychotics in treatment-resistant schizophrenia. Several other uncontrolled historical comparator studies have been conducted on clozapine (Exhibit VII-3).

      The retrospective historical comparator studies in Exhibit VII-3 compare one year before and up to 2.5 years after index treatment with clozapine. Overall results showed a decrease in number of patients hospitalized and/or the hospitalization rate. Results varied from cost increases to cost savings. Several studies demonstrate that although total medical costs may initially increase, by the second year of treatment cost savings begin to be realized.

      There are many limitations to these studies. For example, few studies include a control group. In addition, completer analyses fail to acknowledge cases which dropped out but incurred high medical costs.⁵⁰

References

Where to?

Top of Page
Table of Contents

Home Pages:
Office of Health Policy (HP)
Assistant Secretary for Planning and Evaluation (ASPE)
U.S. Department of Health and Human Services (HHS)

Last updated August 20, 2000